Analysis of community interest: AI benchmarks

In the present analysis we will study the interest of the AI community on a AI benchmarks. We focus on “interest” rather than “progress” for AI benchmarks as this is something we can compute using some proxies. In this particular case, we use the number of (normalised) hits (number of documents) obtained from AItopics per benchmark over the last decade (2008-2019). Note that the results from 2019 are incomplete.

The benchmarks in the present analysis rely on our own previous analysis and annotation of papers, as well as on open resources such as Papers With Code, including data from from several repositories (e.g, EFF, NLP-progress, SQuAD, RedditSota, etc.).

prepareVis <- function(data, rangeMean = 5, norm = T, years = 2008:2019) {

  
  set.seed(288)
  interest.df <- data
  
  ini <- which(colnames(interest.df) == paste0("X",years[1]))
  fin <- which(colnames(interest.df) == paste0("X",years[length(years)]))
  colnames(interest.df)[ini:fin] <- as.character(2008:2019)
  
  
  r <- 1:((length(years)-rangeMean-1))
  years.range <- years[-r]
  interest.df = colwise(type.convert)(interest.df)
  interest.df$mean.Interest <- rowSums(select(interest.df, as.character(years.range)))/(rangeMean+1)
  
  
  # filter(interest.df, is.na(R2))
  # filter(interest.df, R2=="ERROR")
  # # interest.df <- filter(interest.df, !is.na(R2))
  
  
  interest.df$category  <- NA
  for(i in 1:nrow(interest.df)){
    interest.df$category[i] <- str_trim(str_split(interest.df$TaskHierarchies[i], pattern = ">")[[1]][1])
  } # unique(interest.df$category)
  
  
  cogAbs <- c("MP", "SI", "VP", "AP", "AS", "PA", "CE", "CO", "EC", "NV", "CL", "QL", "MS", "MC")
  
  interest <- select(interest.df, one_of(c("keyword", "category",cogAbs,"mean.Interest")))
  keywords <- interest$keyword
  categories <- interest$category
  rownames(interest) <- keywords
  
  
  interest <- interest[,-(1:2)]
  interest["ILSVRC","QL"] <- 0 # <------------------- check it!
  interest = colwise(type.convert)(interest)
  rownames(interest) <- keywords
  interest[interest$mean.Interest ==0, "mean.Interest"]<- 0.0000000000001
  
  
  interest.mean <- interest$mean.Interest
  
  
  if(norm){
    # interest.mean <-  (interest.mean-min(interest.mean))/(max(interest.mean)-min(interest.mean))
    interest <- select(interest, -mean.Interest)
    interest.mean.norm <- normalize(interest.mean+0.000001, method = "scale", range = c(0,1))
    range(interest.mean.norm)
    interest.pond <- interest * interest.mean.norm
    return(list(interest.pond, interest, interest.mean, interest.mean.norm))
    
  }else{
    interest <- select(interest, -mean.Interest)
    interest.pond <- interest * interest.mean
    return(list(interest.pond, interest, interest.mean, interest.mean))
  }
  
}
plotVis <- function(data, categories, norm = T){
    set.seed(288)


  # shapes =  c("square", "triangle", "box", "circle", "dot", "star",
  #             "ellipse", "database", "text", "diamond", "square", "triangle","box")
  # vis$nodes$shape <- c(shapes[as.numeric(as.factor(categories))], rep("#dot",14))
  
  colours = c("1" = "blalck", "2" = "#543005","3" = "#8c510a","4" = "#bf812d",
              "5" = "#dfc27d","6" = "#f6e8c3","7" = "#f5f5f5","8" = "#c7eae5",
              "9" = "#80cdc1", "10" = "#35978f", "11" = "#01665e", "12" = "#003c30", "13" = "#FAFAFA")
  
  vis <- toVisNetworkData(graph_from_incidence_matrix(data, directed = F, weighted = T))
  
  vis$nodes$value = c(rep(10, nrow(vis$nodes)-14), colSums(data)*10000)
  vis$nodes$title <- vis$nodes$label
  vis$nodes$category <- c(categories, rep("CogAb", 14))
  vis$nodes$group <- vis$nodes$category 
  vis$nodes$color <- colours[as.numeric(as.factor(vis$nodes$category))]
 
  # vis$edges$value <- log(vis$edges$weight+1)
  vis$edges$value <- log(normalize(vis$edges$weight+0.00001, method = "range", range = c(0,1))+0.00001)
  
  # vis$edges$width <- vis$edges$weight

  v <- visNetwork(vis$nodes, vis$edges,  height = "1000px", width = "100%") %>% 
    visEdges(arrows = "to", color = list(color = 'rgba(70,130,180,0.3)', highlight ="#4682B4")) %>%
    visIgraphLayout(
      physics = F,
      randomSeed = 2017,
      layout = "layout_with_fr"
    ) %>%  
    visInteraction(navigationButtons = TRUE) %>% 
    visOptions(selectedBy = "group",highlightNearest = TRUE )
  
  return(v)
  
}

interest.df <- read.xlsx2("interest_kw_processed_raw_slope.xlsx", sheetIndex = 1)

df.interest <- data.frame(Benchmark = rownames(prepareVis(interest.df, length(years)-1, norm = F)[[1]]),
                          Last.Decade = prepareVis(interest.df, length(years)-1, norm = F)[[3]], 
                          Last.Lustrum = prepareVis(interest.df, 5, norm = F)[[3]], 
                          Last.Year = prepareVis(interest.df, 1, norm = F)[[3]])

df.interest.m <- melt(df.interest, id.vars = "Benchmark")

a <- ggplot(df.interest.m, aes(reorder(Benchmark,value), value, colour = variable)) + 
  geom_point(alpha = 1/3, size = 3.5) + ylab("Mean Interest") + ylab("") + 
  coord_flip() + theme_minimal() + theme(legend.position="bottom")

b <- ggplot(df.interest.m, aes(reorder(Benchmark,value), log(value), colour = variable)) + 
  geom_point(alpha = 1/3, size = 3.5) + ylab("log(Mean Interest)") + ylab("") + 
  coord_flip() + theme_minimal() + theme(legend.position="bottom")

a
b

interest.df$category  <- NA
for(i in 1:nrow(interest.df)){
  interest.df$category[i] <- str_trim(str_split(interest.df$TaskHierarchies[i], pattern = ">")[[1]][1])
}

categories <- interest.df$category


interest.pond.All <- prepareVis(interest.df, length(years)-1, norm = F)
plotVis(interest.pond.All[[1]], categories)

interest.pond.5 <- prepareVis(interest.df, 5, norm = F)
plotVis(interest.pond.5[[1]], categories)

# barplot(colSums(interest.pond.All[[1]]), main = "Total sum (pondered by mean interest): Last Decade ") # interest.sumcols
# barplot(colSums(interest.pond.5[[1]]), main = "Total sum (pondered by mean interest): Last lustrum") # interest.sumcols


dataALL <- interest.pond.All[[1]]
dataALL$period <- "2008-2019"
dataALL$benchmark <- rownames(dataALL)
meltALL <- melt(dataALL, id.vars = c("benchmark", "period"))


data5 <- interest.pond.5[[1]]
data5$period <- "2014-2019"
data5$benchmark <- rownames(data5)
melt5 <- melt(data5, id.vars = c("benchmark", "period"))


data3 <- prepareVis(interest.df, 3, norm = F)[[1]]
data3$period <- "2017-2019"
data3$benchmark <- rownames(data3)
melt3 <- melt(data3, id.vars = c("benchmark", "period"))


# data1 <- prepareVis(interest.df, 1, norm = F)[[1]]
# data1$period <- "2019"
# data1$benchmark <- rownames(data1)
# melt1 <- melt(data1, id.vars = c("benchmark", "period"))

all <- rbind(meltALL,melt5,melt3)


all.s <- summarise(group_by(all, benchmark, period, variable), mean = mean(value))
ggplotly(ggplot(all.s, aes(variable,mean, fill = period)) + geom_bar(stat = "identity",position = "dodge") + xlab("") + ylab("Mean Interest") + 
           scale_fill_brewer(palette = "Paired") + theme_minimal())

plotIterest.Cat(interest.m, "Computer Vision")

plotIterest.Cat(interest.m, "Graphs")

plotIterest.Cat(interest.m, "Natural Language Processing")

plotIterest.Cat(interest.m, "Playing Games")

plotIterest.Cat(interest.m, "Miscellaneous")

plotIterest.Cat(interest.m, "Medical")

plotIterest.Cat(interest.m, "Methodology")

plotIterest.Cat(interest.m, "Speech")

plotIterest.Cat(interest.m, "Reasoning")

plotIterest.Cat(interest.m, "Time Series")

plotIterest.Cat(interest.m, "Computer Code")